Cell counts about tangram HOT 2 CLOSED

Hi @spatts14 !

This is not a silly question in my opinion, as it is not evident from the tutorial. I am not a developer of the package, but I came up with a workaround.

from anndata import AnnData
from pandas import DataFrame

def get_tangram_predictions(adata: AnnData, perc=0, inplace=True) -> AnnData | DataFrame:
    tg_ct_pred = adata.obsm["tangram_ct_pred"]
    tg_ct_pred = tg_ct_pred.clip(
        tg_ct_pred.quantile(perc), tg_ct_pred.quantile(1 - perc), axis=1
    )
    tg_ct_pred = (tg_ct_pred - tg_ct_pred.min()) / (tg_ct_pred.max() - tg_ct_pred.min())

    df = DataFrame()
    df["tangram_prediction"] = tg_ct_pred.idxmax(axis=1)
    df["tangram_score"] = tg_ct_pred.apply(lambda x: max(x), axis=1)

    if inplace:
        adata.obs = adata.obs.join(df)
        return adata

    return df

This function is inspired by the construct_obs_plot function, which is hidden inside of the plot_cell_annotation_sc function, which plots the predicted annotations.

If you pass inplace=True, then the function will populate adata.obs with two new columns, called "tangram_prediction" and "tangram_score". Otherwise the function will return a dataframe with those columns.

Hope this helps,
Vasily

from tangram.

Thanks so much for your feedback! Sorry for the late response I somehow missed this. Really appreciate your detailed work around @tsvvas

from tangram.